def summary_stats_only_finite(array, alpha=0.05):
""" This version of summary stats only considers values that aren't nan or
inf. Creates summaries a single location based on bootstrap
Args:
array (np.array): np.array that is 1 x n (n is the number of bootstrap
replicates)
Returns:
summary_stats (np.array): np.array that is 1 x 9
pos 0 average
pos 1 min_ci at alpha
pos 2 max_ci at alpha
pos 3 variance
pos 4 least extreme value
pos 5 median
pos 6 median absolute deviation
pos 7 number of infinite values in bootstrap
pos 8 number of nans in bootstrap
"""
finite_vals = np.isfinite(array)
num_inf = np.sum(np.isinf(array))
num_nan = np.sum(np.isnan(array))
if np.sum(finite_vals) == 0:
return (np.nan, np.nan, np.nan, np.nan, np.nan, np.nan, np.nan,
num_inf, num_nan)
else:
these_stats = credible_interval(array[finite_vals], alpha)
this_lev = least_extreme_value(these_stats)
var = np.var(array[finite_vals])
# median = np.median(array[finite_vals])
mad = np.median(np.abs(array[finite_vals] - these_stats[3]))
return (these_stats[0], these_stats[1], these_stats[2], var, this_lev,
these_stats[3], mad, num_inf, num_nan)
def summary_stats_nan_zeros(array, alpha=0.05):
""" This version of summary stats turns nans to 0. NOTE THAT THIS MODIFIES
THE ARRAY
Creates summaries for a single location based on bootstraps
Args:
array (np.array): np.array that is 1 x n (n is the number of bootstrap
replicates)
Returns:
summary_stats (np.array): np.array that is 1 x 9
pos 0 average
pos 1 min_ci at alpha
pos 2 max_ci at alpha
pos 3 variance
pos 4 least extreme value
pos 5 median
pos 6 median absolute deviation
"""
array[np.isnan(array)] = 0
these_stats = credible_interval(array, alpha)
this_lev = least_extreme_value(these_stats)
var = np.var(array)
median = np.median(array)
mad = np.sum(np.abs(array - median)) / array.size
return (these_stats[0], these_stats[1], these_stats[2], var, this_lev,
median, mad)
